Simulation Results and Discussion

First, flow patterns generated in the vicinity of the ridge considered in the present study are described (Fig.4). The qualitative behaviors of the flows numerically simulated by RIAM-COMPACT® and observed in a wind tunnel experiment resemble each other quite closely. As was the case in the previous study²⁾, the shear layer which separated from the vicinity of the ridge top rolls up into isolated vortices (indicated by arrows in Fig.4 (b)). These isolated vortices then form into large-scale vortices, which are periodically shed downstream of the ridge. Refer to Uchida et al.²⁾ for detailed comparisons between the results of the

(a) Wind tunnel experiment, smoke-wire method

(b) Numerical simulation (RIAM-COMPACT®), passive particle tracking

Fig.4 Visualization of the flow field in the vicinity of the ridge; instantaneous field

Prism layer setting thickness h × 8 % number of

layers 25

Flow

(h/U), An examination of Figs. 5 to 7 reveals, in contrast to the previous study²⁾, no significant difference between the results from the simulation which used the steady RANS model (Case 1) and those from the RIAM-COMPACT®

simulation, which used the Smagorinsky LES model (Case 5), indicating that the overall trends of the flow field are similar between the two cases.

The previous study²⁾ investigated airflow around a three-dimensional isolated-hill, and thus, the investigated flow field was complex with the co-existence of various kinds of airflow including flow over the hill and flow around the hill. Probably because of this complex flow field, significant differences arose between the results from the simulations which used the RANS models (i.e., the steady and unsteady RANS models) and those from the simulations which used the LES models (i.e., the standard Smagorinsky and WALE models).

Since numerical simulations performed in the wind power industry are frequently for three-dimensional complex turbulent flow fields, the errors in the mean wind velocity predicted using the RANS models in the previous study²⁾ would greatly affect assessments of the power to be generated and other related variables. Therefore, it can be claimed that the use of the LES models is effective for simulating highly three-dimensional complex turbulent flow fields.

Although not shown here, as was the case in the previous study²⁾, the vertical profiles of the standard deviation of the streamwise (x) wind velocity component evaluated from the results of the simulations in the present study which used the RANS models (i.e., the steady and unsteady RANS models) showed no significant non-zero values at any position. As for the results of the simulations which used the LES models, those from STAR-CCM+ (the standard Smagorinsky and WALE models) and those from RIAM-COMPACT® (the standard Smagorinsky model) showed nearly identical trends

RIAM-COMPACT® natural terrain version of airflow around an isolated-hill with a steep slope angle were compared to those from another commercially-available CFD software package (STAR-CCM+).

In the present paper, in order to examine the prediction accuracy of RIAM- COMPACT® (turbulence model: the standard Smagorinsky LES model), developed by the author's research group, simulations were performed with RIAM-COMPACT® and STAR-CCM+ for airflow over a ridge. The shape of the ridge is identical to the shape created by extending the central cross-section of the isolated-hill from the previous study²⁾ continuously in the spanwise (y) direction. The simulation results were then compared.

For the simulations with STAR-CCM+, both RANS and LES turbulence models were adopted. Specifically, for the RANS turbulence model, two such models were selected for use: the Spalart-Allmaras one-equation eddy-viscosity model (steady RANS) and the SST k-ω two-equation eddy-viscosity model (unsteady RANS). Similarly, for the LES turbulence model (SGS model), two such models were selected for use: the standard Smagorinsky model and the WALE model.

The comparisons of the simulation results revealed the following findings. The significant differences which were found for certain flow characteristics between the results of simulations with the RANS models (the steady and unsteady RANS models) and those with the LES models (the Smagorinsky and WALE models) in the previous study²⁾ were not found in the present study. The flow fields simulated with the use of the RANS models and those simulated with the use of the LES models showed similar overall trends.

The previous study²⁾ investigated airflow around a three-dimensional isolated-hill, and thus, the investigated

(a) Case 1, Simulation results from STAR-CCM+ with the use of the steady RANS model, Spalart-Allmaras one-equation eddy-viscosity turbulence model

(b) Case 5, Simulation results from RIAM-COMPACT® with the use of the LES model, the standard Smagorinsky model, time-averaged field

Fig.5 Comparison of the streamwise (x) velocity component distribution, central plane (y = 0) normal to the spanwise (y) axis.

Here, the velocity component is normalized by the magnitude of the uniform inflow wind velocity.

Flow

Height h

-2h 0h +4h +8h

(a) Case 1, Simulation results from STAR-CCM+ with the use of the steady RANS model, Spalart-Allmaras one-equation eddy-viscosity turbulence model

(b) Case 5, Smulation results from RIAM-COMPACT® with the use of the LES model, the standard Smagorinsky model, time-averaged field

Fig. 6 Comparison of velocity vectors, central plane (y = 0) normal to the spanwise (y) axis.

Here, the velocity components are normalized by the magnitude of the uniform inflow wind velocity.

Height h

-2h 0h +4h +8h

Fig. 7 Comparison of the mean streamwise (x) velocity profiles. Here, the velocity component is normalized by the mean upper-air streamwise (x) velocity above the location of each profile evaluation.

x=-2h x=0h

x=+4h x=+8h

RIAM-COMPACT®

STAR-CCM+

Local increase in velocity Nearly the same values

Similar trends between the profiles from the simulations with RANS and those with LES

Reverse flow region

Similar trends between the profiles from the simulations with RANS and those with LES

for simulating highly three-dimensional complex turbulent flow fields.

As was the case in the previous study²⁾, the vertical profiles of the standard deviation of the streamwise (x) wind velocity component evaluated from the simulations which used the RANS models (i.e., the steady and unsteady RANS models) showed no significant non-zero values at any position in the present study. As for the results of the simulations which used the LES models, the trends of this variable from STAR-CCM+ (the standard Smagorinsky and WALE models) and those from RIAM-COMPACT® (the standard Smagorinsky model) were nearly identical.

Regarding the computational time, although the number of grid points used for the simulation with RIAM-

Wind-Tunnel Experiment–, Reports of RIAM, Kyushu University, No.147, pp.7-14, 2014

2) Takanori UCHIDA, Reproducibility of Complex Turbulent Flow Using Commercially-Available CFD Software –Report 1: For the Case of a Three- Dimensional Isolated-Hill With Steep Slopes–, Reports of RIAM, Kyushu University, No.150, in press, 2016

Appendix

In this appendix, comparisons are made between the results of the simulation with RIAM-COMPACT® (turbulence model: the standard Smagorinsky LES model) from the present study and those from the wind tunnel experiment performed by the author previously. In both Figs.8 and 9, the variables on the horizontal and vertical axes are normalized by the mean upper-air streamwise (x) velocity, Uref, above the location of each profile evaluation and the height of the ridge, h, respectively. The variable z* on the vertical axis represents the height above the flat surface.

The simulations in the present study are performed for airflow over a ridge, the shape of which is identical to the shape created by extending the central cross-section of the isolated-hill from the previous study²⁾ continuously in the spanwise (y) direction.

Therefore, both the simulation and wind tunnel results are more subject to the effect of the blockage ratio than those in the previous study. Specifically, at the ridge top (Fig.8(b)), the wind speed-up ratio for the simulation is significantly overestimated with respect to that for the wind tunnel. As a result, differences in the flow field arise between the two results, including a difference in the size of the vortex region downstream of the ridge (Figs.8(c), 8(d), 9(c), and 9(d)). Although the above-mentioned differences in the flow fields exist, the overall trends of the flow fields are in agreement between the simulated and wind-tunnel flows in Figs.8 and 9.

(a) x=-2h (b) x=0h (c) x=+4h (d) x=+8h Fig.8 Comparison of the mean streamwise (x) velocity profiles,

filled circle: wind tunnel experiment, line: simulation with RIAM-COMPACT® (turbulence model: LES)

(a) x=-2h (b) x=0h (c) x=+4h (d) x=+8h Fig.9 Comparison of the standard deviation of the streamwise (x) velocity,

filled circle: wind tunnel experiment, line: simulation with RIAM-COMPACT® (turbulence model: LES)

non-stratified airflow past a three-dimensional cube were performed. The analysis primarily focused on airflow characteristics in the wake region. The results from the simulation with RIAM-COMPACT® were compared to those from a commercially-available CFD software package (STAR-CCM+) and were found to be in good agreement.

Key words: Commercially-available CFD software, STAR-CCM+, RIAM-COMPACT®, Cube model

1. Introduction

The author's research group has developed the numerical wind diagnosis technique named RIAM-COMPACT® ^{1, 2)}. The core technology of RIAM-COMPACT® is under continuous development at the Research Institute for Applied Mechanics, Kyushu University, Japan. An exclusive license of the core technology has been granted by Kyushu TLO Co., Ltd. (Kyushu University TLO) to RIAM-COMPACT Co., Ltd. (http://www.riam-compact.com/), a venture corporation which was founded by the author and originated at Kyushu University in 2006. A trademark, RIAM-COMPACT®, and a utility model patent were granted to RIAM-COMPACT Co., Ltd. in the same year. In the meantime, a software package has been developed based on the above-mentioned technique and is named the RIAM-COMPACT® natural terrain version software. Efforts have been made to promote this software as a standard software package in the wind power industry.

In the previous paper ^{1, 2)}, a numerical simulation was performed for airflow around an isolated-hill with a steep slope angle using RIAM-COMPACT® natural terrain version, and the results were compared to those from numerical simulations performed using another commercially-available CFD software package.

In the present paper, a similar study is conducted for airflow around a three-dimensional cube,and the results of the comparisons are discussed.

2. Summary of Commercially-Available CFD Software

Commercially-available computational fluid dynamics (CFD) software packages have been developed and used mainly as design tools primarily in the automobile and aviation industries up to the present time. The following is a list of the major CFD software packages available on the market:

General-purpose CFD thermal fluid analysis software packages

■STAR-CCM+

http://www.cd-adapco.co.jp/products/star_ccm_plus/index.

html

■ANSYS(CFD, Fluent, CFX)

http://ansys.jp/solutions/analysis/fluid/index.html

■SCRYU/Tetra

http://www.cradle.co.jp/products/scryutetra/

■STREAM

http://www.cradle.co.jp/products/stream/index.html

■CFD2000

http://www.cae-sc.jp/docs/cfd2000/index.htm

*1 Research Institute for Applied Mechanics, Kyushu University

■PHOENICS

http://www.phoenics.co.jp/

■Autodesk Simulation CFD http://www.cfdesign.com/

■CFD++

http://bakuhatsu.jp/software/cfd/

■CFD-ACE+

http://www.wavefront.co.jp/CAE/cfd-ace-plus/

■AcuSolve

http://acusolve.jsol.co.jp/index.html

■FLOW-3D

http://www.terrabyte.co.jp/FLOW-3D/flow3d.htm

■FloEFD

http://www.sbd.jp/product/netsu/floefd3cad.shtml

■Flow Designer http://www.akl.co.jp/

■PowerFLOW

http://www.exajapan.jp/pages/products/pflow_main.html

■KeyFlow

http://www.kagiken.co.jp/product/keyflow/index.shtml

■OpenFOAM

http://www.cae-sc.jp/docs/FOAM/

■FrontFlow

http://www.advancesoft.jp/product/advance_frontflow_red/

The wind power industry has independently developed and distributed CFD software designed for selecting sites appropriate for the installation of wind turbines (see the list below). Recently, some of the above-listed general-purpose thermal fluid analysis software packages have also started being adopted in the wind power industry.

CFD software packages designed for the wind power industry (wind farm design tools)

■RIAM-COMPACT®

http://www.riam-compact.com/

■MASCOT

http://aquanet21.ddo.jp/mascot/

■WindSim

http://www.windsim.com/

■METEODYN http://meteodyn.com/

In the present paper, the simulation results from RIAM-COMPACT® natural terrain version are compared to those from STAR-CCM+, one of the leading commercially- available CFD software packages. The results of the comparison are discussed.

3. Summary of STAR-CCM+ Software

In this section, a summary of STAR-CCM+, a general-purpose thermal fluid analysis software package distributed by CD-adapco is provided (see Table 1). The version of the software package used in the present study is 8.02.008 (for 64-bit Windows).

STAR-CCM+ uses a single graphical user interface (GUI) for computational grid generation, execution of fluid analyses, and data post-processing. The grid generation method in STAR-CCM+ is distinctive. In STAR-CCM+, both a polyhedral grid and a prism layer grid can be used (for example, see Fig.3). The polyhedral grid is a new type of grid offered and promoted by CD-adapco and consists of polyhedral cells which possess 10 to 15 faces on average.

The use of this cell type makes it possible to dramatically reduce 1) the number of grid cells required to obtain analysis results equivalent to those that can be obtained using a conventional tetrahedral grid and 2) the memory required by the solver. With the use of this cell type, the computational stability improves significantly, and the time required to obtain convergent solutions also decreases. The prism layer grid is a refined grid designed to capture the behavior of the boundary layer that develops over the surface of an object. In this type of grid, layers of thin grid cells are distributed regularly over the object. Since the thickness and number of layers in the normal direction with respect to the object surface can be freely adjusted, the behavior of the boundary layer in the vicinity of a wall can be captured with high accuracy. However, when the number of prism layer grid cells is very large, the computation time increases significantly.

parameters, and simulation set-up used for one of the simulations performed with STAR-CCM+ in the present study, namely the simulation with a steady RANS turbulence model, as an example.

Simulation code STAR-CCM+ v.8.02.008 Governing equation Three-dimensional unsteady

Navier-Stokes equation

Turbulence model

Steady RANS (Spalart-Allmaras one-equation eddy-viscosity

turbulence model)

Time marching

1. First-order implicit unsteady analysis 2. Steady analysis (The steady-state solution is

obtained by specifying the number of time steps.)

Duration of simulation

1. Spin-up: 0 - 100s in non-dimensional time (Time averaging: 100 - 200s

in non-dimensional time) 2. Spin-up: 0 - 2000 in time step number (Time averaging: 2000 - 4000 in time step number)

Discretization of the convective term

A second-order upwind scheme (No options available other than first-order and

second-order upwind schemes)

Gas Constant density

Density ρ 1.0 [kg/m³]

terrain version, developed by the author's research group, will be described. In this software package, a collocated grid in a general curvilinear coordinate system is adopted in order to numerically predict local wind flow over complex terrain with high accuracy while avoiding numerical instability. In this collocated grid, the velocity components and pressure are defined at the grid cell centers, and variables that result from multiplying the contravariant velocity components by the Jacobian are defined at the cell faces. The numerical technique is based on the finite-difference method (FDM), and an LES model is adopted for the turbulence model. In the LES model, a spatial filter is applied to the flow field to separate eddies of various scales into grid-scale (GS) components, which are larger than the computational grid cells, and sub-grid scale (SGS) components, which are smaller than the computational grid cells. Large-scale eddies, i.e., the GS components of turbulence eddies, are directly numerically simulated without the use of a physically simplified model. In contrast, dissipation of energy, which is the main effect of small-scale eddies, i.e., the SGS components, is modeled according to a physics-based analysis of the SGS stress.

For the governing equations of the flow, a spatially-filtered continuity equation for incompressible fluid (Eq.(1)) and a spatially-filtered Navier-Stokes equation (Eq.(2)) are used.

i i

u 0 x

 

 ^-(1)

2 ij

i i i

j

j i j j j

u u p 1 u

t u x x Re x x x

    

    

      ^-(2)

' ' ' '

ij i j k k ij SGS ij

u u 1u u 2 S

  3    -(3)

Fig.1 Computational domain, coordinate system, boundary conditions, and other related information

 

²

SGS C fs s S

   -(4)



ij ij



^1/2

S  2S S -(5)

i j ij

j i

u u S 1

2 x x

  

   

-(6)

 

fs 1 exp z / 25 ^ -(7)



^{h h hx y z}



^1/3

  -(8)

For the computational algorithm, a method similar to a fractional step (FS) method is used, and a time marching method based on the Euler explicit method is adopted. The Poisson’s equation for pressure is solved by the successive over-relaxation (SOR) method. For discretization of all the spatial terms except for the convective term in Eq.(2), a second-order central difference scheme is applied. For the convective term, a third-order upwind difference scheme is applied. An interpolation technique based on four-point differencing and four-point interpolation by Kajishimais used for the fourth-order central differencing that appears in the discretized form of the convective term. For the weighting of the numerical diffusion term in the convective term discretized by third-order upwind differencing, α = 3.0 is commonly applied in the Kawamura-Kuwahara scheme.

However, α = 0.5 is used in the present study to minimize the influence of numerical diffusion. For LES subgrid-scale

modeling, the standard Smagorinsky model is adopted with a model coefficient of 0.1 in conjunction with a wall-damping function.

5. Flow Field and Simulation-setup Considered in the Present Study

In this section, the flow field, coordinate system, and simulation-setup considered for the present study are described (Fig.1).

For the boundary conditions, uniform inflow conditions, free-slip conditions, and convective outflow conditions are applied at the inflow, lateral and upper, and outflow boundaries, respectively. On the surfaces of the cube and the ground surface, no-slip conditions are imposed. In the present study, the Reynolds number is Re (=Uh / ν) = 10⁴, where h is the length of a side of the cube (height) and U is the wind speed at height h at the inflow boundary. The time step is set to Δt = 2.0

Fig.2 Computational grid in the vicinity of the cube from the simulation with RIAM-COMPACT®, structured

grid, central plane (y = 0) normal to the spanwise (y) axis Upper and lateral boundaries: free-slip condition

Inflow boundary:

uniform inflow condition

Ground surface:

no-slip condition

Outflow boundary:

convective outflow condition

h

h

(a) Side view (y = 0), overall view

(b) Side view (y = 0), enlarged view

× 10^-3 (h / U) for the simulation with RIAM-COMPACT®. In contrast, the time step is set to Δt = 2.5 × 10^-2 (h / U) for the simulation with STAR-CCM+.

Fig.2 shows the computational grid (structured grid) used for the simulation with RIAM-COMPACT®. The number of grid points used for this simulation are 326 (x) × 226 (y) × 67 (z) points (approximately 5 million points in total). The grid points in the x- and y-directions are spaced at an even interval of 0.04h, and the grid points in the z-direction are spaced at uneven intervals ranging from 0.003h to 0.6h.

For comparison, Fig.3 shows the computational grid (unstructured grid) used for the simulation with STAR-CCM+.

The total number of grid points is approximately 1.5 million (1/3 of that used for the simulation with RIAM-COMPACT®).

The grid resolution in the vicinity of the cube in the simulation with STAR-CCM+ is set to be nearly identical to that with RIAM-COMPACT®.

Tables 2 and 3 list the turbulence models (RANS and LES models) considered in the present comparative study. For convenience, simulations performed with the use of each of the

(c) Top view (z = 0.5h), overall view

(d) Top view (z = 0.5h), enlarged view

models are referred to as Cases 1 to 5. The models considered in the present study are the same as those from the previous study^{1, 2)}. A brief description of the WALE modelin Case 4 is as follows: the WALE model has been

Case 1

Spalart-Allmaras one-equation eddy-viscosity turbulence model:

steady RANS RANS

models

Case 2

SST k-ω two-equation eddy -viscosity model: unsteady RANS

(URANS)

Case 3 Standard Smagorinsky model:

LES LES models

Case 4 WALE model: LES Table 2 Turbulence models used in the simulations with STAR-CCM+

LES model Case 5 Standard Smagorinsky model:

LES Table 3 Turbulence model

used in the simulation with RIAM-COMPACT®

Fig.3 Computational grid used for the simulation with STAR-CCM+, unstructured grid h

h

(a) Side view (y = 0)

(b) Top view (z = 0.5h), where h is the height of the cube Fig.4 Wind tunnel experiment, flow visualization

by the smoke-wire method, instantaneous field designed in such a way that 1) the eddy viscosity coefficient becomes zero in the vicinity of the ground surface without the use of a wall damping function and 2) the eddy viscosity coefficient is not calculated for laminar shear flow.

6. Simulation Results and Discussions

First, flow patterns generated in the vicinity of the cube considered in the present study are described based on photographs acquired during a wind tunnel experiment (smoke-wire method) (Fig.4). Both the vertical and horizontal cross-sections of the flow field (Figs.4(a) and 4(b), respectively) show that the shear layer which separates from the upstream, upper corners of the cube rolls up into isolated vortices. These isolated vortices then form into large-scale vortices, which are periodically shed downstream of the cube and then advance downstream. As a result, a three- dimensional, complex turbulent flow field is generated in the vicinity of the cube.

Figs.5 to 8 show comparisons of the simulation results.

The time-averaged flow field and the turbulence statistics for Case 5 in these figures (RIAM-COMPACT® with the LES model) were evaluated from the time period t = 100 - 200 (h / U), during which the flow field was fully developed.

Subsequently, comparisons of the visualized flow fields in Figs.5 and 6 are discussed. Due to space limitations, these

figures compare only the simulation results from Case 1 (STAR-CCM+ with the steady RANS model; a steady simulation in which the steady-state solution is obtained by using the number of time steps specified in Table 1 for time marching) and Case 5 (RIAM-COMPACT® with the LES model) as an example. An examination of Figs.5 and 6 reveals the following result: although a complex turbulent flow field is generated in the vicinity of the cube as was described for Fig.4, no significant difference exists between the results from the simulation which uses the RANS model (steady) and those from the simulation which uses the LES model (the standard Smagorinsky model). That is, the overall trends of the results are nearly the same between the two cases.

Figs.7 and 8 show comparisons of various turbulence statistics acquired with the use of the turbulence models considered for comparisons in the present study (RANS and LES models) (Tables 2 and 3). The turbulence statistics were evaluated at the positions x = -1h, 0h, +1h, and +3h. The figures also include the turbulence statistics acquired from a wind tunnel experiment previously conducted by the author.

Comparisons of the profiles of the mean streamwise (x) velocity (Fig.7) reveal that the profiles vary at x = +3h in association with the use of different turbulence models, however, they generally resemble one another at the other positions (x = -1h, 0h, and +1h), independent of the turbulence model used. Comparisons of the standard deviation of the streamwise (x) velocity (Fig.8) show that no significant non-zero values were output at any of the positions when using the RANS models, as was the case in the previous study ^{1, 2)}. In contrast, the simulations with the LES models reveal similar trends for the standard deviation of the streamwise (x) velocity for both the simulations with STAR- CCM+ (the standard Smagorinsky model and the WALE model) and with RIAM-COMPACT® (the standard Smagorinsky model).

Regarding the computational time, although the number of grid points used for the simulation with RIAM-COMPACT®

(structured grid, approximately 5 million points) was approximately three times as large as that used for the simulations with STAR-CCM+ (unstructured grid, approximately 1.5 million points), the simulation with RIAM-COMPACT® completed much faster than those with STAR-CCM+, as was the case in the previous study ^{1, 2)}. Flow

(a) Case 1, Simulation results from STAR-CCM+ with the use of the steady RANS model, Spalart-Allmaras one-equation eddy-viscosity turbulence model

(b) Case 5, Simulation results from RIAM-COMPACT® with the use of the LES model, standard Smagorinsky model, time-averaged field

Fig.5 Comparison of the streamwise (x) velocity component distribution, central plane (y = 0) normal to the spanwise (y) axis.

Here, the velocity component is normalized by the magnitude of the uniform inflow wind velocity.

height h

-1h 0h +1h +3h

(a) Case 1, Simulation results from STAR-CCM+ with the use of the steady RANS model, Spalart-Allmaras one-equation eddy-viscosity turbulence model

(b) Case 5, Simulation results from RIAM-COMPACT® with the use of the LES model, standard Smagorinsky model, time-averaged field

Fig.6 Comparison of velocity vectors, central plane (y = 0) normal to the spanwise (y) axis.

Here, the velocity component is normalized by the magnitude of the uniform inflow wind velocity.

Flow

height h

-1h 0h +1h +3h

Fig. 7 Comparison of the mean streamwise (x) velocity profiles.

Here, the velocity component is normalized by the upper-air wind velocity above the location of each profile evaluation.

x=-1h x=0h

x=+1h x=+3h

Local increase in velocity

Nearly the same values

Similar trends between the profiles from the simulations with RANS and those with LES

Variations in profiles associated with the use of different turbulence models

Reverse flow region

Fig.8 Comparison of the standard deviation of the streamwise (x) velocity.

Here, the velocity component is normalized by the upper-air wind velocity above the location of each profile evaluation.

x=-1h x=0h

x=+1h x=+3h

Occurrence of maxima Nearly the same

values (zero)

Similar profile trends among the simulations

with LES models Similar profile

trends among the simulations with LES models RIAM-COMPACT®

STAR-CCM+

packages, STAR-CCM+. For the simulations with STAR- CCM+, both RANS and LES turbulence models were adopted. Specifically, for the RANS turbulence model, two such models were selected for use: the Spalart-Allmaras one-equation eddy-viscosity model (steady RANS) and the SST k-ω two-equation eddy-viscosity model (unsteady RANS). Similarly, for the LES turbulence model, two such models were selected for use: the standard Smagorinsky model and the WALE model.

The comparisons of the simulation results from the use of the four turbulence models revealed the following findings.

Comparisons of the visualized large-scale flow fields showed no significant differences between the simulation results from the use of the RANS turbulence models (i.e., steady and unsteady RANS models) and those from the use of the LES models (i.e., the standard Smagorinsky model and the WALE model) despite the generation of a complex turbulent flow field in the vicinity of the cube. The overall trends of the visualized flow field were nearly the same among all the simulations.

Comparisons of the mean streamwise (x) velocity profiles revealed variations in the velocity at x = +3h, associated with

of grid points used for the simulation with RIAM- COMPACT® was approximately three times as large (structured grid, approximately 5 million grid points) as that used for the simulations with STAR-CCM+ (unstructured grid, approximately 1.5 million grid points), the simulation with RIAM-COMPACT® completed much faster than those with STAR-CCM+, as was the case in the previous study

1,2 ).

References

1) Takanori UCHIDA, Reproducibility of Complex Turbulent Flow Using Commercially-Available CFD Software –Report 1: For the Case of a Three- Dimensional Isolated-Hill With Steep Slopes–, Reports of RIAM, Kyushu University, No.150, in press, 2016 2) Takanori UCHIDA, Reproducibility of Complex

Turbulent Flow Using Commercially-Available CFD Software –Report 2: For the Case of a Two- Dimensional Ridge With Steep Slopes–, Reports of RIAM, Kyushu University, No.150, in press, 2016

ドキュメント内 九州大学応用力学研究所 Reports　No.150（Mar. 2016） (ページ 67-86)